Error in train.py

[jimpap1]

When i run the train.py it fails to load images and i also get TypeErrors.Any thoughts? Did i miss anything?

[ajaysridhar0]

Hi @jimpap1,

Can you explain how you downloaded and processed the datasets? It's hard to pinpoint why the images failed to load without more context.

Thanks,
Ajay

[swx153]

试试把.yaml文件里数据集的绝对路径换成相对路径

[PEACHTTT]

Hi @jimpap1, 你好

Can you explain how you downloaded and processed the datasets? It's hard to pinpoint why the images failed to load without more context.您能解释一下您是如何下载和处理数据集的吗？如果没有更多上下文，很难确定为什么图像无法加载。

Thanks, 谢谢 Ajay 阿杰

@ajaysridhar0 Hello,I met same problems about"Faild to load and TypeError".I download the datasets through the link of README.Then I processed it with data_split.py(I only tried the goStanford datasets),then change the directory in the vint.yaml to run train.py

[PEACHTTT]

试试把.yaml文件里数据集的绝对路径换成相对路径

尝试了一下没有用，请问你是没有遇到这个问题训练代码跑通了吗

[swx153]

试试把.yaml文件里数据集的绝对路径换成相对路径

尝试了一下没有用，请问你是没有遇到这个问题训练代码跑通了吗

我之前遇到了其他问题，试着把这里的路径改了之后，就出现了无法加载图片的问题，然后我又改回去了。我已经跑通了

[PEACHTTT]

@swx153 感谢回复！我还想确认一下关于stanford的那个数据集，是不是只需要data_split.py划分处理。我有搜到pytorch版本问题也会导致这个原因，可以问一下你的torch版本吗？方便的话能不能加一下联系方式，我想对比下是哪一步出了问题。目前我卡在这一步

[PEACHTTT]

@swx153 十分感谢！抱歉我暂时不参与这个相关项目了，谢谢你的耐心回复！

[jimpap1]

Why do i get Aborted (core dumped) when i run train.py ?

[plusgrey]

Why do i get Aborted (core dumped) when i run train.py ?

Same with torch 2.4.0, ubuntu 18.04

[plusgrey]

Why do i get Aborted (core dumped) when i run train.py ?

Same with torch 2.4.0, ubuntu 18.04

I have solved the problem by downgrading the opencv to 4.1.2.30. I think two main reasons will cause the "Aborted (core dumped)" error. 1: Too small image_log_freq (>0) or too large num_image_log, these settings takes lots of memory which might cause this error. 2: In some machine, the version of PyQT5 mismatches the opencv-python version used in ViNT, this will also cause the error. p.s. sometimes, the torch.multiprocessing can also cause this error.

[LeZheng-x]

Maybe you should delete dataset_xx.lmdb file, and try run the train.py again. I found once the detection cache file is created, it will not be updated.You can try to use gdb to output whether _image_cache has reasonable key-value pairs in vint_dataset.py. Hope it will help you!